I second the LES comment, I have very rarely used those models but I've always been told that it's best practice to stick to fine isotropic meshes. What I'm more worried about is your cell count, in the image from paraview you provided, as far as I can see, the mesh is extremely coarse. 1 cm physical size is very large, especially when trying to capture finer detail in a small channel. I would personally try to do a preliminary mesh sensitivity study to see how the result changes with a finer mesh. Maybe the vortex is trying to form, but the large cells simply don't allow that to happen. I'm currently working with T-channel flows and vortex formation and behaviour is extremely sensitive to the mesh, at least in my experience.